A study of ~57,000 kidney disease patients in the Boston area finds that an algorithm used to decide priority for transplants was biased against Black patients
A formula for assessing the gravity of kidney disease is one of many that is adjusted for race. The practice can exacerbate health disparities.
Context & Ripple Effects
This is the second act of a bias story that opened a year earlier, when researchers found an algorithm assessing medical needs for millions of US patients systematically understated the needs of Black patients [[a:947222]]. Two weeks before this study landed, a STAT investigation detailed how software targeting stepped-up care was already infusing racial bias into clinical decision-making [[a:958974]]. The new finding moves the pattern from care management to transplant triage — a higher-stakes setting where miscalculation costs organs, not just attention.
It also complicates the earlier framing of kidney-care AI as unambiguous progress: coverage of donor-matching tools shortening paired-exchange times [[a:933333]] now sits alongside evidence that the severity formulas feeding these systems carry race adjustments that work against the very patients they rank.
First-order effects
- Black patients in the Boston-area system studied had their transplant priority computed by a formula that discounted disease severity, meaning some were ranked lower on waiting lists than their clinical condition warranted.
- Hospitals and clinicians relying on this race-adjusted formula face immediate pressure to re-score or re-validate priority decisions made under it.
Second-order effects
- Vendors and medical bodies behind other race-adjusted calculators come under pressure to audit their own formulas, extending the scrutiny from care-allocation software flagged in the STAT investigation to diagnostic and staging tools themselves.
- The finding reinforces concerns raised about eye-disease AI trained mostly on US, European, and Chinese patient populations [[a:958887]] — datasets and formula design choices that bake disparities in become procurement liabilities for health systems buying these tools.
Third-order effects
- If the pattern holds, clinical algorithms move toward the kind of formal validation regime drug-like interventions face — a trajectory already visible in the UK, where the NHS's National Liver Offering Scheme was found to contain a fatal error while deciding liver transplant eligibility [[a:846139]].
- Health systems may increasingly demand bias audits as a condition of deploying scoring algorithms, structurally shifting power toward independent validators over tool vendors.
The trend: Clinical scoring algorithms are moving from trusted infrastructure to audited infrastructure, with racial bias findings in transplant and care allocation driving validation requirements across medicine.